Drier conditions diminish plant coexistence via soil microbes in a tropical humid forest

Abstract

Premise

Global environmental change can alter ecological mechanisms that maintain biodiversity. Interactions between plants and soil microbes mediate plant species coexistence, which can vary with abiotic factors such as light and soil moisture, and such context dependence is only beginning to be explored.

Methods

We assessed how variation in light and soil moisture alters soil microbial effects on the predicted coexistence of two common tree species (Litsea floribunda and Symplocos racemosa) native to the Western Ghats, India. We conducted a shade‐house reciprocal transplant experiment with different soil origins in factorial combinations of high/low water and light, and predicted coexistence outcomes using metrics that decompose microbial effects into stabilization and fitness differences. For high‐water, low‐light conditions, we also evaluated whether soil microbes alter plant‐plant interactions, using a structural framework that quantifies the feasibility domain of coexistence from species interaction coefficients.

Results

Soil microbes were predicted to promote plant coexistence in high‐water, low‐light conditions, where stabilization exceeded the fitness differences due to microbes. Under low‐water, high‐light conditions, larger fitness differences predicted exclusion of S. racemosa. Microbes had weak effects on plant‐plant interactions, but predicted coexistence improved slightly in soils with background microbes not shaped by either species, due to weaker estimated fitness differences in those soils.

Conclusions

Drier and brighter conditions may weaken the potential for microbially mediated plant coexistence in tropical forests. These findings hint at shifts in microbially mediated plant community dynamics in response to global change factors such as forest fragmentation and drought.

Global environmental change can alter the mechanisms that maintain diversity in ecological communities (Bellard et al., 2012; Haddad et al., 2015). One key process for plant community dynamics involves the interactions between plants and soil microbes, which can mediate coexistence of plant species and hence the maintenance of plant diversity (Bever, 2003). However, microbial effects on plant performance can depend on the abiotic context, as these factors can change what microbes occur and plant response to microbes. It follows that the role of microbes in mediating plant species' coexistence will be context dependent as well, which is increasingly recognized (e.g., Dudenhöffer et al., 2022; Hassan et al., 2022; Xi et al., 2020, 2023). Despite recent efforts to understand context dependence in microbially mediated coexistence, tropical forests remain missing from the picture (Krishnadas et al., 2024). Only with empirical data from different ecosystems can we assess generality in the strength and direction of abiotic modulation of microbially mediated plant coexistence. Moreover, evaluations of context dependence in microbial effects on plant coexistence have not incorporated recent theoretical developments, posing another gap in our understanding of the problem.

The role of microbes in plant coexistence is commonly assessed through stabilizing and destabilizing effects of reciprocal plant‐soil feedback (Bever et al., 1997). The feedback framework evaluates coexistence from the degree to which plant growth is affected by soil microbial communities conditioned by conspecific versus heterospecific plants. Specifically, when plant growth is suppressed or favored more in soils conditioned by conspecific plants than in soils conditioned by heterospecific plants, ensuing negative (stabilizing) or positive (destabilizing) feedback, respectively, can promote or hinder coexistence (Bever et al., 1997; Kandlikar et al., 2019). Recent theory, however, shows that plant species can vary in their average sensitivity to conspecific and heterospecific microbes, resulting in microbially mediated fitness differences (Kandlikar et al., 2019; Kandlikar, 2024). Coexistence between competing species requires stabilizing feedback to outweigh fitness differences (Chesson, 2000; Barabás et al., 2018; Kandlikar et al., 2019; Yan et al., 2022), yet this comparison is missing from many studies of microbially mediated coexistence (Kandlikar, 2024). In this vein, studies of context dependence in microbially mediated plant coexistence have also focused on the stabilization component.

Abiotic conditions could alter both stabilization and fitness differences to affect coexistence (Collings et al., 2025). For example, soil moisture regulates microbial activity and composition as well as their effects on plant growth (Manrubia et al., 2019; Zhang et al., 2024). If more moist soils promote the proliferation of soil pathogens, especially species‐specific ones, wetter conditions would be predicted to yield stronger stabilization and thus promote coexistence (Milici et al., 2025). Alternatively, proliferation of pathogens in moist soils could increase fitness differences if one of the plant species is more susceptible (Xi et al., 2025). Drier conditions may decrease the activity of soil pathogens or increase plant dependence on mutualists, leading to less negative (or more positive) feedback (Snyder and Harmon‐Threatt, 2019; Milici et al., 2020; but see Dudenhöffer et al., 2022). However, drier conditions might increase or decrease the fitness differences between plant species, depending on whether soil microbes favor the weaker or the stronger competitor (Milici et al., 2020; Collings et al., 2025; Xi et al., 2025).

Light is another key variable that affects plant performance and plant‐microbe interactions in closed‐canopy forests (Smith and Reynolds, 2015; Xi et al., 2023). Brighter conditions may enable plants to invest more resources in root exudates to suppress pathogens or cultivate mutualists to improve performance in their own soils (Augspurger, 1984; Martin et al., 2018). Consequently, feedback may become less negative or more positive in brighter conditions (Smith and Reynolds, 2015; Xi et al., 2023). Increased light can also allow plants to more easily replace tissues lost to pathogens, potentially mitigating performance detriments due to pathogens (Augspurger, 1984; Poorter and Kitajima 2007; del Pino et al., 2015). If decreased susceptibility to pathogens disproportionately benefits stronger competitors, it could exacerbate fitness differences between species (Xi et al., 2023). Thus, in brighter conditions, soil microbes may diminish the coexistence of competing plant species.

Additionally, while classical plant‐soil feedback theory is based on microbial impacts on species' intrinsic growth rates, microbes can also affect plant coexistence by altering direct plant‐plant interactions (Lekberg et al., 2018; Ke and Wan, 2020). Plants may experience stronger competition from conspecifics when conditioned soil microbes are largely pathogenic and suppress plant growth (Pendergast et al., 2013), whereas plants that cultivate soil microbial communities dominated by mutualistic microbes can experience weaker intraspecific competition (Wagg and McKenzie‐Gopsill, 2023) or enhanced conspecific growth compared to heterospecifics (Koziol and Bever, 2019), thereby destabilizing coexistence. Recent empirical studies have evaluated the consequences of soil microbes for plant competition (e.g., Van Nuland et al., 2023; Wang et al., 2024), but insights from tropical forests are lacking.

We asked how abiotic context alters the role of soil microbes in plant coexistence in a tropical humid forest and whether microbes alter the outcomes of competitive interactions between plants. Using a shade‐house experiment, we quantified stabilization and fitness differences mediated by soil microbes for seedlings of two tree species, Litsea floribunda (Blume) Gamble (Lauraceae) and Symplocos racemosa Roxb. (Symplocaceae), under factorial combinations of high and low light and water levels. Although microbes can impact plant coexistence by altering various demographic processes (reviewed in Ke et al., 2025), we focused on the seedling stage because soil microbes can have large impacts on plant performance at this stage (Mangan et al., 2010) and because the seedling bank plays a crucial role in maintaining diversity and composition of forests (Green et al., 2014). We hypothesized that soil microbes would promote exclusion in low‐water and high‐light conditions due to increased fitness differences and weaker stabilization. For the role of soil microbes in modifying plant‐plant competition in wet and shaded conditions, we expected that conspecific soils would increase intraspecific competition to a greater extent than interspecific competition.

MATERIALS AND METHODS

We conducted a shade‐house experiment using seedlings of two tree species, Symplocos racemosa (hereafter Symplocos) and Litsea floribunda (hereafter Litsea). Both species are widely distributed and relatively common in the tropical humid forests of the Western Ghats, a global biodiversity hotspot (12°56′ N, 75°39′ E). Both tree species fruit between May and July, with seeds germinating during September and October, the latter part of the monsoon season. We conducted the study in forests of Hassan district, Karnataka, India (Krishnadas et al., 2018). The area receives annual rainfall of 4000 mm, most of which falls in June–October. The soils are primarily clayey Alfisols with good drainage, originating from a gneissic base (Krishnadas et al., 2018). Monthly mean temperatures range from ~20°C in January to ~24°C in April, with maximum temperatures occasionally exceeding 30°C during the pre‐monsoon dry period. Relative humidity follows a complementary seasonal trend, remaining close to saturation (95%–100%) during the monsoon, but falling to 60%–70% during the dry season.

Seed and soil collection

Between June and August 2022, we collected 500–700 seeds from three or four individuals per species within the forest. Seeds were surface‐sterilized in a 0.5% bleach solution for 10–12 min and washed with water before being sown in trays filled with non‐sterilized cocopeat. Trays were placed inside a shaded hall in the field station and misted every other day until germination. After four to five weeks, seedlings were transferred to a 1 L bag of sterilized potting mix (60% soil from nearby forests, 20% sand, 20% cocopeat; autoclaved at 121°C for 20 min cycles) to allow growth while experimental treatments were prepared.

Plants were grown with one of four soil inocula in the experimental treatments. Two treatments comprised soils with microbial communities conditioned by adults of our two focal species. We collected these soils from within a 1 m radius of adult trees (DBH ≥ 10 cm) growing in the interior of the forest, ensuring that there was no adult of the other species within a 10 m radius of the collection points. We cleared the organic litter layer and collected soil from the top 15 cm. Another treatment represented a “forest reference” soil (hereafter “forest soil”) collected from areas of the forest interior without adults of either focal species in a 10 m radius. A final treatment used an autoclaved version of this forest soil, which we refer to as the “sterile reference” (hereafter “sterile soil”). We collected six replicates of each inoculum from the forest and used each replicate to inoculate pots in one block in our experiment. Soils were collected using ethanol‐sterilized equipment and stored for ≤48 h at 4°C before being added to experimental pots.

Experimental design

In late November to early December 2022, we transplanted seedlings into 5 L pots containing 4.5 L of sterilized (autoclaved) potting mix and 500 mL of soil inoculum (10% by volume). Each pot received one of four soil inoculum treatments described above, in separate experiments designed to answer the two focal questions.

Our first goal was to understand how light and soil moisture influence microbially mediated seedling performance and potential coexistence through plant‐soil feedback. To this end, we grew seedlings of each species in the four soil treatments: conspecific, heterospecific, forest soil, and sterile soil, replicated in factorial combinations of high and low levels of light and water (Figure 1A). We used one layer of 75% shade net for the high‐light condition, and two layers to create the low‐light condition. Light levels in high‐ versus low‐light treatments were ~10,000 lux and ~1000 lux, respectively, as measured with a digital lux meter (LX‐103 Light Meter, HTC Instruments, Mumbai, Maharashtra, India) (Appendix S1: Figure S7). We monitored soil moisture levels as volumetric water content using a Hydrosense II soil moisture probe (Campbell Scientific, Logan, Utah, USA). For the high‐water condition, pots were watered twice a week; for the low‐water condition, pots were watered once every 7–10 days, depending on ambient temperature and soil moisture sensor readings (Appendix S1: Figure S6). Light and water treatments began after seedlings were acclimated for 10 days after transplant.

Figure 1.

Experimental design and treatment structure of the shade‐house experiment with Litsea floribunda and Symplocos racemosa. Pot color indicates soil treatment, and seedling color denotes species identity (see key at left). (A) The plant‐soil feedback experiment, conducted under a fully factorial design with two water levels (high/low) × 2 light levels (high/low) across six blocks (192 seedlings). (B) The competition experiment, in which pots contained two seedlings (conspecific or heterospecific pairs). Due to limited seed availability, the experiment was conducted only under high‐water, low‐light conditions (120 seedlings).

Our second goal was to assess how soil microbes influence coexistence through plant‐plant competition—in other words, how soil microbes affect seedling interactions under conditions resembling intact, shaded, and humid forest understories. We addressed this question only in one abiotic condition (low‐light/high‐water) due to limited availability of seedlings. We chose the low‐light/high‐water condition because it represents the baseline for forests. Seedlings of each species were grown with one heterospecific or conspecific neighbor in each of the four soil inoculum treatments (Figure 1B). We did not include treatments where focal seedlings experienced conspecific competition in heterospecific‐conditioned soil, due to limited seedling availability and due to limited biological relevance within our analytical framework (coexistence metrics are evaluated under an invasion logic whereby each species experiences intraspecific competition in its own conditioned soil; for details, see Appendix S1).

After approximately five months of growth, seedlings were harvested, dried in a hot air oven for four to seven days at 60°C, and weighed for aboveground and belowground dry masses in a weighing balance with sensitivity up to three decimal places (ME204, Mettler Toledo, Columbus, Ohio, USA).

Data analysis

Microbial impacts on plant performance

To evaluate how light and water alter microbially mediated seedling growth, we assessed plant performance as the total seedling biomass (aboveground and belowground). We quantified the effects of soil microbes on plant performance as the log biomass ratio in conditioned compared to reference soil (Equation 1).

$$m_{\mathit{ij}}=\log \frac{\text{Biomass of}\mathit{i}\in \text{soil}j}{\text{Biomass of}\mathit{i}\in \text{reference soil}}$$ (1)

Hence, m _LL and m _LS refer to the effects of Litsea‐conditioned and Symplocos‐conditioned soils, respectively, on the growth of Litsea seedlings, while m _SS and m _SL refer to the effects of Symplocos‐conditioned and Litsea‐conditioned soils on the growth of Symplocos seedlings. Using sterile soil as the “reference” condition allows us to estimate the total or net effects of microbes on plant performance, and using forest soil as the “reference” isolates the effects of the species‐specific microbes that are conditioned by either plant species (Yan et al., 2022; Kandlikar, 2024).

We used linear models to evaluate how microbial effects on plant growth varied across the four abiotic conditions in the experiment (factorial combination of high/low light and high/low water). For each species, we fitted the model below (Equation 2) to estimate how m _ij , representing the (log) response of species i to soil conditioned by species j, varied with conspecific and heterospecific soils (soil type) for each environmental condition (a single categorical variable with factorial combinations of light and water treatments). Before fitting the model, we set conspecific soil and the high‐water/high‐light treatment as the reference levels for factor coding. The response variable was the log biomass ratio, and the “−1” in Equation 2 indicates that the model was fit without an intercept. Treatment differences were evaluated using estimated marginal means from the fitted model.

$$m_{\mathit{ij}}=\text{soil} \text{type}\times \text{water}\times \text{light}-1$$ (2)

Preliminary analyses showed negligible variance attributable to blocks, so we proceeded with linear models. From the final models, we evaluated the estimated marginal means using the R package emmeans (Lenth, 2025) to test for significant differences in conspecific and heterospecific microbial effects between light and water conditions. We used Tukey's honestly significant difference (HSD) test to adjust for multiple comparisons.

Coexistence metrics

While the individual m _ij terms capture the effects of soil microbes on seedling growth, evaluating microbially mediated coexistence for a species pair requires comparing the degree to which soils affect the growth of plants from conditioning plant species versus that of the other species to generate microbially mediated (de)stabilization and fitness differences (Kandlikar et al., 2019). The strength of (de)stabilization (Equation 3) is determined by the degree to which each conditioned soil community affects the growth of conspecific versus heterospecific seedlings (Bever et al., 1997). A positive value implies stabilization via negative feedback, while a negative value implies destabilization.

$$\text{Stabilization}=\frac{1}{2}(m_{\text{LL}}-m_{\text{LS}}-m_{\text{SL}}+m_{\text{SS}})$$ (3)

Microbially mediated fitness differences (Equation 4) reflect the average response of each plant species to conditioned soils, compared to a reference soil, which may be either sterile or forest soil.

$$\text{Fitness difference}=\frac{1}{2}(m_{\text{LL}}+m_{\text{LS}}-m_{\text{SL}}-m_{\text{SS}})$$ (4)

Assuming no other competitive differences except those mediated by soil microbes, coexistence requires that stabilization exceeds fitness differences. Conversely, if fitness differences exceed stabilization, microbes are predicted to drive exclusion of the weaker species. Priority effects arise if microbes drive stronger destabilization than fitness differences between the focal plant species (Kandlikar, 2024). We also evaluated microbial impacts on predicted coexistence dynamics by calculating each species' low‐density growth rate, which reflects each species' ability to invade the other species' resident population (we present detailed results of this analysis only in Appendix S1, as they exactly complement the stabilization/fitness difference comparisons in Figure 2).

Figure 2.

Effects of light and water availability on the impacts of conspecific and heterospecific soil microbes on plant performance in Litsea floribunda and Symplocos racemosa. For each parameter m _ij , the y‐axis shows the log biomass ratio with (A) forest‐collected soil or (B) autoclaved forest soil used as reference. A value of zero (dashed line) indicates equal performance in conditioned soil compared to the reference soil. Positive values indicate that the focal species performs better in the conditioned soil than in the reference soil, whereas negative values indicate reduced performance in conditioned soil. Points show estimated mean log biomass ratios, error bars show 95% confidence intervals from the linear regression model, and asterisks denote significant differences between treatment pairs (Tukey‐adjusted p < 0.05).

Interaction coefficients and structural coexistence metrics

Our second question focused on the role of soil microbial communities on predicted plant coexistence by mediating seedling interactions. While metrics of niche and fitness differences from classical coexistence theory largely assume competitive interactions between species (Bimler et al., 2018; Chesson, 2018; Ke and Wan, 2020), many of the plant‐plant interactions we observed in our experiment were positive (i.e., seedling growth was often higher when grown with a neighbor than without). We therefore adopted a structural approach to quantify microbial impacts on coexistence potential (Saavedra et al., 2017). This approach characterizes the feasibility domain, the set of intrinsic growth rate combinations that allow all species to maintain positive equilibrium abundances, by evaluating the geometric constraints that competition coefficients impose on coexistence (Saavedra et al., 2017).

The details of this analysis are presented in Appendix S1. Briefly, we estimated initial biomass using species‐specific allometric relationships, then used a Bayesian approach to fit a modified Ricker model to infer each species' intrinsic growth rates (λ) and interaction coefficients (α) across soil microbial treatments (brms: Bürkner, 2017; Stan: Carpenter et al., 2017; cmdstanr: Ricker, 1954; Gabry et al., 2024). Posterior samples of λ and α were assembled into treatment‐specific interaction matrices (posterior: Bürkner et al., 2025), including a derived “conditioned soil” scenario to reflect invasion into resident‐conditioned microbial environments. We then used the structural framework of Saavedra et al. (2017) to translate these posterior matrices into geometric coexistence metrics. Specifically, we quantified the posterior distribution of the feasibility domain (Ω), which represents a structural analogue of stabilization, as well as the alignment of observed growth rates with that domain (Θ), which is a structural analogue of the fitness difference.

All statistical analyses were performed using R version 4.5.0 (R Core Team, 2025) in RStudio version 2024.12.1 (Posit Team, 2025).

RESULTS

How does variation in light and water alter the outcomes of microbially mediated plant coexistence?

We observed context dependence in species‐level biomass responses to soil microbes. Microbial effects on Symplocos biomass varied with abiotic context, particularly under high‐light conditions where conspecific microbes enhanced growth by 64% in high‐water compared to low‐water soils (Figure 2B). This shift was less pronounced when measured against forest soil, suggesting that background microbial communities affect species performance in addition to species‐specific microbes (Figure 2A). In low‐light conditions, Symplocos showed minimal responses to soil microbes regardless of water availability, indicating light as a key modulator of plant‐microbe interactions for this species. By contrast, growth of Litsea showed little response to soil microbes across all light and water combinations (Figure 2A, B).

We used the joint effects of species' biomass responses (m_ij ) (Equations 3 and 4) to predict how variation in light and water alters microbially mediated plant coexistence. Regardless of whether we used live or sterilized forest soil as the reference, predicted coexistence was more likely under high‐water, low‐light conditions (55%–66% of simulated outcomes; leftmost column in Figure 3A, B). Species exclusion was the most likely outcome predicted for other conditions, particularly low‐water, high‐light conditions (44%–86%). Specifically, in low‐water conditions, soil microbes provided a fitness advantage to Litsea, with this advantage most pronounced under high‐light (Figure 3A, B). Only under low‐water, high‐light conditions did the two reference soils differ in most common outcomes (sterilized forest soil: exclusion of Symplocos in 86%; live forest soil: coexistence in 47% of outcomes; Appendix S1: Table S1), revealing that microbial effects on coexistence differ between background soil communities and plant‐conditioned microbes (Figure 3A, B). Both reference soils yielded similar predicted coexistence outcomes in high‐water, low‐light conditions, but there was greater uncertainty in outcomes when forest soil was the reference, wherein Symplocos had a higher chance of being excluded. High‐water, high‐light conditions represented the sole environment where Symplocos gained a fitness advantage, with exclusion of Litsea predicted in 38%–40% of simulation outcomes and with priority effects also common (31%; Figure 3A, B).

Figure 3.

Effects of light and water availability on soil microbe‐mediated coexistence outcomes in Litsea floribunda and Symplocos racemosa: (A) predicted coexistence outcomes of conditioned microbes compared to live forest soils and (B) these outcomes compared to sterile soils. Each panel represents a factorial combination of high or low water and high or low light. Stabilization (Equation 3) reflects the extent to which soil microbial communities generate negative frequency dependence, with positive values promoting coexistence. For fitness differences (Equation 4), positive values favor Litsea and negative values favor Symplocos. Shaded areas delineate zones corresponding to different competitive outcomes (coexistence, priority effects, or competitive exclusion). Large black points in each panel represent the mean values of stabilization and fitness differences, and clouds of small points were constructed from a simulation‐based resampling approach reflecting the uncertainty in the mean estimates (see text for details).

Across treatments, fitness differences were substantially more sensitive to abiotic context than stabilization (Appendix S1: Table S10). When evaluated in comparison to the sterile reference, fitness differences varied by 0.95 units across environments (calculated as the maximum minus minimum mean value), nearly 1.8× the corresponding variation in stabilization (SD = 0.53; Appendix S1: Table S10). Similarly, variation among environments, quantified as the standard deviation across treatments, was higher for fitness differences (SD = 0.40; Appendix S1: Table S10) than for stabilization (SD = 0.23; Appendix S1: Table S10). Stabilization peaked under high‐water, low‐light conditions (mean stabilization = 0.44; Appendix S1: Table S1) and became negative (i.e., destabilizing) under high‐water, high‐light conditions (mean stabilization = −0.10).

Do soil microbes modify plant‐plant competition in high‐water, low‐light conditions?

To examine whether soil microbes modify plant‐plant competition in the high‐water/low‐light conditions, we fit a Bayesian competition model to predict seedling biomass and evaluated both model predictive performance and posterior shifts in key parameters (Figure 4; Appendix S1: Figures S2 and S3 and Tables S5–S9). Intrinsic growth rates (λ) overlapped broadly among microbial treatments for both species (Appendix S1: Figure S2 and Table S5). Model comparison provided limited support for soil effects on competition coefficients; specifically, although the model allowing soil effects on competition coefficients had the best numerical predictive performance, its advantage over simpler models was small and uncertain (ΔELPD = 1.02, SE = 4.3; Appendix S1: Figure S3 and Table S8). However, intraspecific competition shifted from near‐zero or negative values in sterilized soil (α _SS = − 0.233; α _LL = 0.024; Figure 4; Appendix S1: Table S5) to positive values in forest soil (α _SS = 0.50, Pr(>0) = 0.98; α _LL = 0.34, Pr(>0) = 0.93; Figure 4; Appendix S1: Table S5). These estimated shifts in competition were also associated with higher predicted coexistence potential in forest soil, with structural feasibility increasing from 1.5% (sterilized) to 57.7% (forest) and driven primarily by reduced structural fitness differences (θ decreasing from 106° to 23°) rather than large increases in structural niche differences (Ω increasing from 0.20 to 0.43). Full model structure, parameter estimates across all soil treatments (including species‐conditioned soils), and additional diagnostics are provided in Appendix S1 (Figure 4; Appendix S1: Figures S2–S3 and Tables S5–S9).

Figure 4.

Posterior distributions of structural coexistence metrics and competition coefficients across soil microbial treatments in Litsea floribunda and Symplocos racemosa. (A, B) Posterior density distributions of structural niche differences (Ω) and structural fitness differences (Θ, in degrees). (C–F) Posterior distributions of competition coefficients (α) describing intraspecific and interspecific interactions between Litsea and Symplocos. Density curves are derived from posterior draws of the Bayesian competition model. Colors indicate soil microbial treatments (forest soil, Litsea‐conditioned soil, Symplocos‐conditioned soil, and sterilized soil). Positive α values indicate stronger net competitive effects, whereas negative values indicate net facilitative effects. Treatments lacking empirical data for a specific interaction (e.g., conspecific interaction coefficients in heterospecific‐conditioned soil) were excluded.

DISCUSSION

For two locally abundant co‐occurring tree species from the Western Ghats of India, Litsea floribunda and Symplocos racemosa, we found that species‐specific soil microbes were predicted to promote species coexistence most strongly in high‐water and low‐light conditions. Although microbial effects on plant growth in relation to background soils were not prominent (Figure 2), microbial contributions to plant growth combined to shift the joint distributions of stabilization and fitness differences, producing detectable changes in predicted outcomes for coexistence (Figure 3). The shift in predictions of microbially mediated coexistence across abiotic contexts occurred through changes in fitness differences more than stabilization, suggesting a pathway by which environmental changes (e.g., drought or forest fragmentation) could alter plant coexistence. Finally, although we did not find evidence for strong impacts of soil microbes on species interactions in this study, our analysis of the impacts of soil microbes on structural niche and fitness differences suggests that microbes from background forest soils may contribute to coexistence.

The two focal tree species differed markedly in their responses to soil microbes across abiotic contexts. For Symplocos, species‐specific soil microbes drove strong context‐dependent responses: under high‐light conditions, microbial effects shifted from positive in high‐water conditions to strongly negative under low‐water (Figure 2). By contrast, Litsea maintained relatively consistent performance across microbial treatments and abiotic conditions (Figure 2). Low‐water conditions may have compromised plant performance directly but differentially for the two species—Symplocos seedlings showed greater wilting in low‐water treatments, although we did not formally quantify wilt level—and drought effects may have intensified under high‐light conditions through increased transpiration demand (Choudhury et al., 2017). Reduced moisture may also have altered plant‐microbe interactions, potentially including mycorrhizal associations, although these were not directly measured in our study. If confirmed, such effects would be consistent with evidence from tropical and subtropical systems that reduced soil moisture can lower arbuscular mycorrhizal colonization and restructure root‐associated fungal communities (Lozano et al., 2021; Zhang et al., 2024). An ongoing sequencing study in this system indicates that both Symplocos and Litsea harbor mycorrhizal fungi (V. P. Saini et al., unpublished data). Differences between species in harnessing mutualists in high‐ versus low‐water conditions may have contributed to the differences in microbial effects on predicted coexistence in those conditions.

Biomass responses to microbes across abiotic contexts propagated to alter predicted coexistence outcomes. Microbially mediated coexistence was predicted to be the most likely outcome in low‐light, high‐water conditions, whereas exclusion of Symplocos was predicted to be the most likely outcome in high‐light, low‐water conditions. This shift accrued primarily through amplification of microbially mediated fitness differences, with comparatively smaller changes in stabilization, possibly reflecting greater sensitivity of Symplocos to pathogens and/or disruption of mycorrhizal associations in low‐water conditions. Our results differ from those of Pan et al. (2024), who found that negative feedback from conspecific soil microbes weakened in high‐light and low‐water conditions, which may be due to a stronger influence of species‐specific microbes compared to our system.

When biomass responses to species‐specific soils were evaluated using forest soils as the reference, the effects of species‐specific microbes became less clear, and fitness differences appeared to decline. High uncertainty in coexistence estimates when using live forest soil as the reference indicates that for these two species, under the conditions tested here, effects of plant‐specific microbes can be weakened by background microbial communities (Kandlikar, 2024). The difference in predicted outcomes with sterile versus live reference soils is consistent with either a gain of beneficial microbes from live forest soil or differences between the two species in their responses to non‐host‐specific pathogens (Kandlikar, 2024). Thus, light and water may change coexistence outcomes by influencing whether background and host‐specific soil microbes favor the weaker or the stronger competitor (Semchenko et al., 2022; Pan et al., 2024; Xi et al., 2025). Both species were common in this forest, which may reflect general tolerance of pathogenic microbes and/or their ability to harness mutualists.

Our findings point to possible shifts in microbially mediated coexistence of plant species in human‐modified forests. The abiotic treatments in this study reflect abiotic contrasts observed in the landscape between forest edges that are often drier and brighter than shaded, moist interiors (Camargo and Kapos, 1995; Heithecker and Halpern, 2007). Previous work in this system found that stabilizing effects of soil microbes during seed‐to‐seedling transition can weaken at forest edges (Krishnadas et al., 2018; Krishnadas and Comita, 2019). Ongoing work at our site also indicates that negative effects of conspecifics on seedling survival weakens in drier, brighter conditions (R. Jhaveri et al., unpublished data). Our results further indicate that soil microbes may increase fitness differences in the drier, brighter conditions typical of forest edges. Furthermore, that soil moisture modulates these coexistence components is consistent with recent evidence that negative density dependence intensifies in wetter years (Lebrija‐Trejos et al., 2023). Thus, edge effects on plant‐microbe interactions may be a potential pathway for changes in plant community dynamics in fragmented forests.

Our results suggest that soil microbes can alter the nature of seedling interactions, but there was substantial uncertainty in the outcomes. When seedlings were grown with a neighbor, background soil microbes were associated with lower estimated fitness differences. However, due to weak support for soil effects on the competition coefficients, we cannot confidently attribute those shifts in coexistence potential to microbial changes in plant‐plant competition. This contrasts with the results of Wang et al. (2024), who found that soil microbes reduced predicted coexistence by increasing fitness differences more than niche differences. Whether soil microbial effects on seedling interactions also vary with abiotic context should be tested, as suggested by recent observational studies (Lebrija‐Trejos et al., 2023). While we primarily observed neutral or positive interactions between seedlings in our study, evaluating seedling interactions at higher densities could yield more insight into the role of soil microbes in shaping density‐dependent demographic performance of seedlings (Ke and Wan, 2023), observed here and in other tropical forests.

CONCLUSIONS

We examined how plant coexistence mediated by soil microbes varies with abiotic context and plant competition. Our results suggest that conditions of low water and high light, compared to high water and low light, can increase microbially mediated fitness differences and thereby increase the likelihood of species exclusion. This pattern indicates that mechanisms maintaining plant diversity through soil microbes may weaken as abiotic conditions shift, such as with edge effects in human‐modified forests or drought due to global change. However, our conclusions are based on coexistence predicted from seedling performance and model‐based metrics, whereas realized long‐term coexistence can also be shaped by demographic processes operating later in life. Also, our study was limited to two species, and the outcome of microbially mediated coexistence may differ with species life history and traits, especially traits that directly influence response to soil microbes (Rutten and Allan, 2023). That microbial effects on coexistence could shift with competitors suggests a need to explore soil microbial effects on seedling interactions under changing abiotic conditions. Overall, our results underscore the dynamic nature of plant‐soil microbe interactions, with potential fallouts for tropical forest diversity in an era of rapid environmental change.

AUTHOR CONTRIBUTIONS

V.P.S., M.K., and G.S.K. collaboratively designed research questions and experimental design; V.P.S. conducted fieldwork and experiments with assistance from G.S.K. and M.K.; V.P.S. analyzed the data with review and editing by M.K. and G.S.K.; V.P.S. wrote the first draft of the manuscript with substantial inputs from M.K. and G.S.K.; M.K. and G.S.K. provided critical revisions and editorial contributions.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Appendix S1: Supplementary methods, figures, and tables.

DATA AVAILABILITY STATEMENT

All the data and code supporting the analysis for this paper are archived on Zenodo and publicly available at: https://doi.org/10.5281/zenodo.19757765.